In-Vehicle Object Presence Detection Using Multi-Sensor Confidence Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle monitoring systems face challenges in directly detecting the presence of a child or animal in the cabin, especially when there is limited visibility from camera sensors, necessitating a method to facilitate detection in various situations.
Innovation Solution
A computer-implemented method that processes data from multiple in-vehicle devices to accumulate scores indicative of the presence or absence of an object, determining a confidence value based on these scores, and comparing it to a threshold to detect the object's presence, using machine-learning models and incorporating parameters from image, audio, and user profile data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If camera sensors are used to directly detect objects in the cabin, then detection accuracy is improved, but the number of sensors and system complexity increase
Solution Approach 1:
The system segments the detection task across multiple data sources rather than relying on a single sensor type. Image data, audio data, and user profile data are processed separately and then integrated, allowing each data source to contribute to the overall detection accuracy without requiring every sensor to be present
Solution Approach 2:
The system introduces an intermediary processing layer that combines indirect indicators (audio cues, user profiles, image patterns) to infer object presence. This intermediary approach allows detection without direct visual confirmation, reducing the need for multiple camera sensors while maintaining detection capability
2Adaptability or versatility
If radar sensors are used to detect through materials, then detection capability is improved, but device complexity and cost increase
Solution Approach 1:
The system makes existing multi-functional sensors (cameras, microphones) perform additional detection functions by analyzing their data for indirect signs of object presence. These sensors already exist for other purposes, but are now also used for object detection through alternative data interpretation
Solution Approach 2:
Instead of using radar sensors that can detect through materials, the system creates a virtual model of object presence by copying and integrating information from multiple indirect sources (audio, images, profiles). This virtual detection model achieves similar versatility without requiring physical radar hardware
3Area of stationary object
If multiple camera sensors are positioned to cover the full cabin, then detection coverage is improved, but device complexity and installation difficulty increase
Solution Approach 1:
The system adds temporal and data-type dimensions to the detection process. Instead of expanding spatial coverage with more cameras, it processes data over time and integrates multiple data types (image, audio, profile) to achieve comprehensive detection from limited camera positions
Solution Approach 2:
The system performs preliminary analysis of user profiles and historical data to predict likely object locations and characteristics. This preliminary action allows the system to focus camera monitoring on high-probability areas, effectively expanding detection coverage without adding sensors to low-probability zones
4Reliability
If direct visibility from image sensor is required, then detection reliability is improved, but detection capability in obscured situations deteriorates
Solution Approach 1:
The system uses feedback from multiple data sources to compensate for lack of direct visual confirmation. Audio feedback, user profile feedback, and temporal pattern feedback create a feedback loop that maintains detection reliability even when direct camera visibility is blocked
Solution Approach 2:
Instead of requiring direct visual evidence to prove object presence, the system inverts the logic by detecting the absence of expected indicators (no audio, no profile match, no indirect visual cues) to infer absence. This inversion allows reliable detection in obscured situations by what is missing rather than what is present
Data Source
AI summary
Systems and techniques are described herein for detecting a state of presence of a given object in a vehicle. In aspects, techniques include processing data received over time from various devices in the vehicle to determine a set of accumulated scores. The data includes parameters indicative of a presence or absence of a detected object in the vehicle. Further, the processing may be effective to determine a set of accumulated scores. The techniques further include determining an object-related confidence value representing a likelihood that the detected object is present in the vehicle, based on the set of accumulated scores, and then comparing the object-related confidence value to a predetermined threshold. The comparison is sufficient to detect the presence of the detected object in the vehicle if the object-related confidence value exceeds the predetermined threshold.